n8n Agents

The n8n AI Agent node (@n8n/n8n-nodes-langchain.agent) is a multi-turn LLM driver with sub-nodes for the model, memory, tools, and an optional output parser. This skill is the deep guide to designing agents and the LangChain family around them. For the high-level "where an agent fits in a workflow" picture, see n8n-workflow-patterns ai_agent_workflow.md — this skill goes one level down into how to build it well.

For node-type formats: in workflow JSON the LangChain nodes use the long @n8n/n8n-nodes-langchain.* form (.agent, .lmChatOpenAi, .memoryBufferWindow, .outputParserStructured, .toolWorkflow, .toolHttpRequest, .toolCode). When you call get_node / validate_node, use the short form (nodes-langchain.agent). See n8n-mcp-tools-expert for the format rules.

Pick the right node first

Reaching for an Agent when the task is one-shot classification or extraction is the most common over-build. Decide before you wire anything:

Text Classifier detail (the Agent + Switch anti-pattern): every category needs both a name AND a description. The model routes against the description, not the name — a category with no description gets picked by coin-flip. Set options.enableAutoFixing: true for robustness on edge inputs. One node, N branches, done. Reaching for an Agent that "decides" then a Switch that "routes" is two nodes plus prompt boilerplate for what Text Classifier does natively.

Chat-model nodes (.lmChatOpenAi, .lmChatAnthropic, .lmChatOpenRouter, …) are sub-nodes — they don't run standalone. They wire into a chain, agent, classifier, or extractor via the ai_languageModel connection.

The sub-node pattern

The Agent has a main input (the prompt / user message) and up to four sub-node slots, each wired by its own ai_* connection type:

A sub-node connects FROM itself TO the agent. In workflow JSON the connection lives on the sub-node, keyed by the ai_* type:

"Main LLM": {
  "ai_languageModel": [[{ "node": "AI Agent", "type": "ai_languageModel", "index": 0 }]]
},
"Simple Memory": {
  "ai_memory": [[{ "node": "AI Agent", "type": "ai_memory", "index": 0 }]]
},
"Search customer DB": {
  "ai_tool": [[{ "node": "AI Agent", "type": "ai_tool", "index": 0 }]]
}

Multiple tools all connect into the same ai_tool index 0 — they stack, they don't fan into separate indices. With n8n_update_partial_workflow you wire each with an addConnection op using sourceOutput: "ai_tool". The agent puts its final answer in $json.output (not .text, not .response) — downstream nodes read {{ $json.output }}.

See EXAMPLES.md for a complete stateless agent-core node-object snippet.

Two non-negotiables

1. Tool names and descriptions ARE part of the prompt. The model picks a tool by reading its name and description — nothing else. A tool named tool1 with an empty description is invisible to the model: it skips it, mis-selects it, or hallucinates parameters. There's usually no error — just an agent that "won't use my tool". Treat both like API design. → TOOLS.md 2. Structured output must parse AND autoFix. An outputParserStructured with autoFix: true and a coding-capable fixer model is the production pattern. Without autoFix, one malformed JSON response halts the whole workflow. → STRUCTURED_OUTPUT.md

Strong defaults

• Per-tool usage goes in the tool description, not the system prompt. Anything about how to call this specific tool belongs with the tool, so it travels across agents and keeps the system prompt focused. → SYSTEM_PROMPT.md
• Sub-workflow tools (.toolWorkflow) for anything multi-step. Any workflow becomes a tool with typed $fromAI() inputs, and composes with branching, error handling, and reuse. Default here when in doubt. → SUBWORKFLOW_AS_TOOL.md and n8n-subworkflows.
• Wrap tools with user-visible side effects in human review. Sends, payments, refunds, account changes get gated behind an approval node so a human signs off before the tool fires. → HUMAN_REVIEW.md
• Raise maxIterations. The default tool-call cap is low (single digits on most versions) — fine for a one-tool agent, far too low for a multi-tool agent that chains several calls per turn. It surfaces as "max iterations reached" or empty output. Set options.maxIterations to a realistic ceiling (15 for a focused sub-agent, 50-200 for a broad orchestrator).
• Put the current date in the system prompt via {{ $now }} (or {{ $now.format('DDDD') }}). A hardcoded date is stale immediately.

The four tool types

Pick the lightest option that covers the job:

There is also a Custom Code Tool (.toolCode) for pure inline computation — but its runtime contract (string in / string out, no $fromAI, no $helpers) is owned by the n8n-code-tool skill. Read that before writing one. Rule of thumb: if you find yourself reaching for $fromAI() inside the code, you want .toolWorkflow instead.

$fromAI(): how the agent fills tool parameters

Tool parameters the agent should decide are wrapped in $fromAI(). It is a real n8n expression helper, used inside a tool node's parameter expressions:

={{ $fromAI('paramName', 'what to put here — be specific: format, range, example', 'string') }}

• paramName — the name the model uses internally (snake_case or camelCase, be consistent).
• description — tells the model what value to produce. It is part of the prompt — write it like JSDoc.
• type (optional) — 'string' (default), 'number', 'boolean', 'json'. A wrong-typed value fails the call.
• defaultValue (optional) — used when the model omits it.

$fromAI() carries JSON only — it cannot carry binary (no base64, no file bytes). And not every parameter has to be $fromAI: plumb identity, authority limits, and correlation IDs (userId, refund caps, sessionId) deterministically from workflow context so the agent can't get them wrong or even see them. → TOOLS.md for the full anatomy and the "give the agent a button, not a steering wheel" pattern.

System prompt vs tool description

Why split it: a well-described tool works in any agent that drops it in, tool details only "load" when the model considers that tool (token efficiency), and you update one tool description instead of a paragraph buried in a 5000-token prompt. → SYSTEM_PROMPT.md

Structured output: when and how

Add an outputParserStructured sub-node (wired ai_outputParser) when downstream needs strict JSON, not free-form text. Two rules:

1. Use schemaType: 'manual' with a real JSON Schema, not jsonSchemaExample. An example can't express required-vs-optional, enums, numeric ranges, or array constraints — you outgrow it the first time the shape gets non-trivial. Reach for fromJson + an example only for throwaway shapes. 2. autoFix: true with a coding-capable fixer model. Wire a second model into the parser's ai_languageModel slot. Reconciling broken JSON against a schema is a coding task — a weak fixer just produces another malformed retry and burns tokens.

→ STRUCTURED_OUTPUT.md for the schema patterns, the load-bearing "DO NOT wrap in markdown" retry line, and the parse-failure cookbook.

Memory: brief mental model

Memory is a sub-node (ai_memory). Without it, every call is stateless — correct for one-shot tasks (classify, summarize). With it, the agent holds a conversation, keyed by whatever expression you bind to sessionKey.

• memoryBufferWindow — keeps the last N exchanges per key and persists across executions via n8n's store. The default for chat. contextWindowLength defaults to 5, which is very low — 50 is a saner starting point. Messages past the window are gone entirely.
• memoryPostgresChat / memoryRedisChat — only when memory must be read outside the agent (your own UI, analytics, cross-system). Not needed just to survive restarts; BufferWindow already does that.

Plumb a stable key from the trigger to memory consistently. Chat triggers fill sessionId automatically; for other surfaces derive one (Slack thread_ts, a webhook conversation ID). Never hardcode sessionId: 'default' and never put sessionId behind $fromAI (the model will fabricate a UUID). → MEMORY.md

Binary and the agent boundary

This is the seam that trips people up:

• The model CAN see uploaded images (vision) via options.passthroughBinaryImages: true on the agent.
• Tools CANNOT receive binary. $fromAI() is JSON-only — no base64, no bytes, even through non-AI bindings.
• The agent's output is text-shaped (or structured-text with a parser). When a model returns image/audio/video bytes, the Agent doesn't surface them at all — there's nothing to recover downstream.

Workaround: pre-stage uploads to storage before the agent runs, inject the storage keys into the system prompt, and let tools accept the key as a string parameter and re-fetch internally. For one-shot media generation, skip the agent and call the provider's native single-call node directly.

The binary mechanics (which storage, how to stage, how to re-fetch) are owned by n8n-binary-and-data — see its agent-tool binary reference. This skill only marks the boundary; don't re-derive the mechanics here.

Human review (gate destructive tools)

When a tool's effect needs human sign-off before execution (sends, payments, refunds, account changes), wrap it with a review tool node — slackHitlTool, discordHitlTool, telegramHitlTool, gmailHitlTool, etc. (n8n names these "Hitl" / human-in-the-loop). The review node sits between the wrapped tool and the agent on the ai_tool connection: wrapped tool → review node → Agent.

Whether sign-off is needed is a product/policy call — surface the question to the user, recommend based on blast radius, and let them decide.

The critical rule: show the actual parameters the wrapped tool will receive. Use the literal {{ $tool.parameters.<name> }} in the approval message, never a $fromAI() paraphrase — otherwise the human approves text the model made up, not the call about to fire. → HUMAN_REVIEW.md

Chat agents (Slack, Discord, Teams, Telegram)

The one non-negotiable, regardless of complexity: any chat-triggered workflow that posts a reply MUST filter out the bot's own user ID, or its own replies re-trigger it in an infinite loop that burns runs and tokens. Prefer trigger-level filtering when available (Slack Trigger's options.userIds is an exclusion list — put the bot ID there); otherwise filter $json.user !== '<BOT_USER_ID>' in the first node after the trigger.

Beyond the filter, a simple bot (trigger → agent → reply) lives fine in one workflow. Split into shell + core + sub-agents only once you need loading UX, sub-agents, multi-surface reuse, or robust error handling:

• Shell — trigger, anti-loop filter, event-type Switch, loading/error UX, renders the reply. No LLM.
• Core — stateless agent, chatInput + threadId inputs, memory keyed on threadId, tools and sub-agents.
• Sub-agents — one narrow domain each, called via .toolWorkflow, stateless (full context in chatInput).

→ CHAT_AGENT_PATTERNS.md for per-surface semantics, threading-as-session, and the full topology.

Persisted n8n Agents (n8n_manage_agents)

A persisted n8n Agent is a different artifact from the AI Agent node covered above: a standalone assistant record — model, instructions, tools, skills, tasks, memory, channels — stored and versioned by n8n itself, managed through n8n_manage_agents (n8n's instance-level MCP server), not a node inside a workflow's JSON.

Prerequisites: N8N_MCP_ACCESS_TOKEN configured (separate from the Public API key) and n8n 2.34+ with the agents module enabled. The token is required for every action, including reference/search — without it, nothing works. Separately, reference and search work for any agent regardless of MCP exposure; every other action needs the target agent exposed to MCP (agents created through this tool are exposed automatically — the exposure gate only matters for agents that already existed before this tool touched them).

Build sequence: 1. action: "reference" — read the config schema and the exact mutate operations before anything else. 2. action: "discover_assets" — list what the agent can actually be wired to. Takes projectId (from n8n_list_catalog({kind: "projects"})) and kind: models (with a provider), integrations, workflows, subagents or mcpServers. One call per kind. 3. action: "create" — projectId, name, config. 4. action: "mutate" — one resource per call (config.patch, skill.upsert/delete, task.upsert/delete, customTool.upsert/delete), always carrying the latest hash forward. Mind the two names: n8n returns it as configHash and expects it back as args.baseConfigHash. args are forwarded to n8n verbatim, so a near-miss on any field name comes back as INVALID_ARGS, not a helpful correction — which is why step 1 reads the schema first. A stale hash comes back as STALE_CONFIG — re-get and retry with the fresh one. 5. action: "validate" — before offering to call or publish. 6. action: "publish" — only on the user's explicit request, never proactively.

action: "call" runs the agent with real credentials and real tools — a live execution, not a dry run. A result can carry approvals[] for tool calls that need a human decision; never approve on the user's behalf — surface them and resume only after the user decides.

Custom tools are a third code runtime — don't reuse either of the others. A customTool.upsert body is TypeScript, and the only imports it may use are @n8n/agents and zod. This is not the Code node (JavaScript/Python, returns [{json: …}]) and not the AI-agent Custom Code Tool covered by n8n-code-tool (@n8n/n8n-nodes-langchain.toolCode, returns a string, no $fromAI()). Reaching for the wrong contract is the easy mistake here, because all three are "write code the agent calls". Read the shape from action: "reference" before writing one; a compile failure or an unknown agentId surfaces as AGENT_TOOL_ERROR.

Credential caveat: on n8n 2.36.x the agents runtime rejects azureOpenAiApi and aws credentials (reported as missing: ["credential"]); the response's hint names the accepted types instead.

Testing without leaving debris: name throwaway agents [TEST] … and delete them when you're done — a persisted Agent outlives the conversation that made it, unlike a workflow you can leave inactive.

→ n8n-mcp-tools-expert ## Agents for the tool's full action list and error codes.

RAG (retrieval augmented generation)

n8n ships the LangChain RAG primitives (document loaders, splitters, embeddings, vector stores, retrievers). Two opinions worth stating up front:

1. Rule out cheaper lookups first. Exact lookups → a database or Data Table query, not RAG. Freshness → a live search tool. A small/structured doc set → give the agent list/fetch tools. Reach for a vector store only when there are too many docs to list and queries are semantic. 2. Wire the vector store as a retrieval tool (mode: 'retrieve-as-tool', ai_tool) so the agent decides when retrieval is relevant and can phrase the query itself. Embed query and documents with the same model.

→ RAG.md (intentionally thin — defaults depend on data shape and scale).

Reference files

Anti-patterns

What's NOT available via the community MCP

What the MCP can do: search and inspect every LangChain node (search_nodes, get_node), validate node config and the whole graph (validate_node, validate_workflow), build and patch the agent and its sub-nodes (n8n_update_partial_workflow with addConnection on ai_* outputs), test (n8n_test_workflow), and pull the saved JSON to verify wiring (n8n_get_workflow). The deep AI-agent guide also lives in tools_documentation({topic: "ai_agents_guide", depth: "full"}).

Integration with other skills

• n8n-workflow-patterns (ai_agent_workflow.md) — the high-level "agent in a workflow" shape. This skill is the deep dive; start there for architecture.
• n8n-mcp-tools-expert — node-type formats (short form for get_node, long form in JSON) and tool-selection guidance. Consult before any MCP call.
• n8n-node-configuration — displayOptions-driven fields on the agent and sub-nodes; Slack/Block Kit message shapes (NODE_FAMILY_GOTCHAS.md, Slack section).
• n8n-expression-syntax — {{ }}, $json.output, $now, and $fromAI/$tool.parameters all rely on correct expression syntax.
• n8n-code-tool — the Custom Code Tool's runtime contract (string in/out, no $fromAI). Read it before writing a .toolCode.
• n8n-subworkflows — the sub-workflow primitive that .toolWorkflow builds on (Execute Workflow Trigger inputs/outputs, naming, search-before-build).
• n8n-binary-and-data — owns the agent-tool binary boundary mechanics (staging uploads, returning generated files).
• n8n-validation-expert — interpreting validate_workflow results, including AI-connection issues (a tool wired into main instead of ai_tool flags as disconnected).
• n8n-error-handling — onError: 'continueErrorOutput' on tool sub-workflows and the agent-core call; error UX on chat shells.
• n8n-code-javascript / n8n-code-python — for Code-node logic inside a tool sub-workflow (different sandbox from the Code Tool).

Quick reference checklist

Before shipping an agent:

• Right node: Agent for tools/memory/multi-turn; Text Classifier for routing; Information Extractor for fields; native node for media
• Model wired via ai_languageModel
• Every tool has a verb-first specific name AND a real description
• $fromAI() descriptions are specific (format, range, example); identity/limits/sessionId plumbed deterministically, not via $fromAI
• Per-tool guidance lives in tool descriptions, not the system prompt
• $now in the system prompt (no hardcoded date)
• maxIterations raised for multi-tool agents
• Memory keyed on a stable sessionKey from the trigger (not 'default', not $fromAI); contextWindowLength raised from 5
• Structured output: schemaType: 'manual' + autoFix: true + a coding-capable fixer model
• Destructive tools wrapped in human review; approval message uses $tool.parameters, not $fromAI
• Chat bots filter the bot's own user ID (trigger-level or first node)
• Binary: model vision via passthroughBinaryImages; tools get storage keys, never bytes
• Validated with validate_workflow and verified with n8n_get_workflow (sub-nodes on ai_*, not main)

Remember: an agent is only as good as its tool names, descriptions, and system-prompt discipline. The model can't see your wiring — it sees a system prompt and a list of named, described tools. Design those like an API and most "the agent won't behave" problems disappear.